Tencent opens Hy4 flash weights and adapts it to MetaX GPUs
AI summary
Tencent Hunyuan has formally opened the model weights for Hy4 flash, according to 36Kr. MetaX said its Xiyun C-series GPUs completed Day 0 end-to-end adaptation for Hy4 flash, with inference validation on the same day. MetaX also claimed the setup can support high-throughput, low-cost, scaled Agent workloads, but no performance figures were provided.
In-depth insight
The open weights and immediate hardware adaptation broaden Hy4 flash’s deployment options, while MetaX’s workload benefits remain vendor claims rather than independently reported results.
沐曦股份曦云C系列GPU完成腾讯混元Hy4 flash适配-36氪
登录
搜索